Description of AESFAST

This model is based on a physical formulation of transmittance, namely the Goody random model.  It can 
therefore work on any vertical coordinate (as opposed to fixed pressure levels).  It is presented in Garand et 
al, 1999.  The model considers up to 8 gases. The forward model is accompanied by an analytic Jacobian 
routine.  Jacobians for temperature and each of the 8 gases  are available (most fast models compute only T, 
H2O and O3 Jacobians).  It can be used in data assimilation for NWP models.  The Goody formulation is 
used for 6 gases: H2O, CO2, O3, CO, CH4 and N2O.  Continua for H2O, O2 and N2 are also considered.   
The optical depths of each gas is multiplied by an adjustment factor, typically closed to unity, for a global 
fit to LBL transmittances available from an ensemble of 189 profiles.  The most dominant gas in the 
channel is adjusted last in such a way as to minimize the errors in terms of resulting brightness temperature 
considering all gases.  The procedure to establish adjustment factors is fully automated, allowing quick 
adaptation to new sets of HIRS-like channels.  The model is designed for infrared channels only (no 
scattering).  It allows variable surface emissivity.  Cloudy atmospheres can be treated as well with the 
effective cloud amount provided in each layer as input.  

Garand, L., D. S. Turner, C. Chouinard, and J. Halle, 1999. A physical formulation of atmospheric 
transmittance for the massive assimilation of satellite infrared radiances. J. Appl. Meteor., 38,  541-554.




Description of AESFAST2

This is the same model as AESFAST, except that the Goody random formulation of transmittance is now 
supplemented by an additional dependency on pressure for the mean line width.  This dependency is of the 
form (1. + c P/Po) where c is the adjustment factor, P is the pressure of the layer and Po is a reference 
pressure.  This new adjustment provided  a very large improvement in HIRS-02 (for CO2) and a significant 
improvement in HIRS-09 (for O3). The minimization procedure to find “c” is straightforward  but we find 
in practice that the end result may be influenced by the initial choice of c in some channels.  With the 
evolution to narrow channels, it is preferable to pre-determine, if possible, areas where the original 
AESFAST can best work.